Biography: Michael Mbagwu, MD, is a practicing comprehensive ophthalmologist at the VA Palo Alto Health Care System and is a current Byers Ophthalmic Innovation Fellow at the Byers Eye Institute at Stanford, Stanford University Medical School. He has a research interest in medical informatics, application of real-world data and creation of AI algorithms for diagnosis/prognosis of ophthalmic disease.
Biography:

Theodore Leng, MD, MS, is a practicing retina specialist, associate professor of ophthalmology and director of clinical and translational research at the Byers Eye Institute at Stanford, Stanford University Medical School. He has a research background in image analysis, machine learning algorithms, and clinical trial design and execution.

Disclosures: Leng reports he serves as a medical advisor to Verana Health. Mbagwu reports he is a part-time consultant at Verana Health.
August 12, 2021
2 min read
Save

BLOG: Linking imaging data with patient profiles in the IRIS Registry

Biography: Michael Mbagwu, MD, is a practicing comprehensive ophthalmologist at the VA Palo Alto Health Care System and is a current Byers Ophthalmic Innovation Fellow at the Byers Eye Institute at Stanford, Stanford University Medical School. He has a research interest in medical informatics, application of real-world data and creation of AI algorithms for diagnosis/prognosis of ophthalmic disease.
Biography:

Theodore Leng, MD, MS, is a practicing retina specialist, associate professor of ophthalmology and director of clinical and translational research at the Byers Eye Institute at Stanford, Stanford University Medical School. He has a research background in image analysis, machine learning algorithms, and clinical trial design and execution.

Disclosures: Leng reports he serves as a medical advisor to Verana Health. Mbagwu reports he is a part-time consultant at Verana Health.
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

The American Academy of Ophthalmology IRIS Registry contains 367 million patient encounters from more than 65 million unique patients.

As the IRIS Registry grows, it could be vital to link images and imaging reports — which are not currently housed in the IRIS Registry — to electronic health record files so researchers using the IRIS Registry could harness its power to extract meaningful insights and inspire innovations.

Linking images and reports to the corresponding EHR clinical notes in the IRIS Registry is not a one-to-one process. The issue lies in metadata, the identifying information such as patient name, sex and date of birth that is linked to an image. Patient demographic data in EHRs may not align with metadata used for categorization and filing in an imaging vendor’s local system. If vendors began aligning metadata to a single standard — at this point, few vendors do so — then broad cross-platform linkage of imaging data to EHR files could be achieved.

Luckily, such a standard exists as the Digital Imaging and Communications in Medicine (DICOM) standard. Although few vendors fully comply to all DICOM standards, eventual uniformity could move imaging data out of its various vendor-based silos and into the IRIS Registry.

Verana Health, the Academy’s data curation and analytics partner, sought to understand the feasibility of connecting vendor-based images with IRIS Registry EHR files. To do so, researchers gathered imaging data from six different posterior segment imaging modalities from two vendors and extracted metadata from those files. After analyzing these metadata, a program was written to harmonize vendor-based metadata to DICOM standards. Imaging files that were able to align with DICOM standards were then linked to IRIS Registry EHR files based on the degree to which they contained matching data.

In some cases, vendors captured metadata that aligned with DICOM requirements. In other cases, some vendors captured metadata that, while useful within a particular vendor’s platform for the purposes of local organization, were not congruent with DICOM standards.

Approximately 1.8 million DICOM files from approximately 58,500 unique patients were acquired. Researchers matched approximately 1.5 million DICOM files and 48,500 unique patients to patients in the IRIS Registry, resulting in a match rate of roughly 83%.

In some cases, imaging metadata did not align with IRIS Registry EHR files. Imaging files used as “test cases,” patient detail mismatch between electronic systems and algorithmic parameter aberrations are all possible reasons why imaging data did not link to a patient profile in the IRIS Registry.

Creating an automated means by which to standardize data from multiple imaging modalities and vendors allows Verana Health to help the Academy grow the IRIS Registry’s scope and depth. Researchers wishing to leverage imaging-based longitudinal real-world data can do so on a large scale when imaging data is linked to EHR files in the IRIS Registry, as it allows researchers to observe anatomic results following particular interventions detailed in patient clinical charts.

In this study, we were able to build more robust profiles for approximately 48,500 patients. Given that the IRIS Registry contains data from 65 million unique patients, there is a long way to go. Still, this research showed that a program written to harmonize vendor-based imaging data with IRIS Registry EHR files is feasible.

Reference:

Mbagwu M, et al. Large scale cross-vendor linkage of ophthalmic imaging and American Academy of Ophthalmology IRIS Registry clinical data. Presented at: Association for Research in Vision and Ophthalmology, May 1-7, 2021 (virtual meeting).